Encoding biological recognition in a bicomponentcell-membrane mimicCesar Rodriguez-Emmeneggera,b, Qi Xiaoc,d, Nina Yu. Kostinaa,b, Samuel E. Shermanc, Khosrow Rahimia,b,Benjamin E. Partridgec, Shangda Lic, Dipankar Sahooc, Aracelee M. Reveron Perezc, Irene Buzzaccheraa,c,e, Hong Hanc,Meir Kerznerc, Ishita Malhotrac, Martin Möllera,b, Christopher J. Wilsone, Matthew C. Goodf,g, Mark Goulianh,i,Tobias Baumgartc, Michael L. Kleind,1, and Virgil Percecc,1

aDWI–Leibniz Institute for Interactive Materials, Rheinisch-Westfälische Technische Hochschule Aachen University, 52074 Aachen, Germany; bInstitute ofTechnical and Macromolecular Chemistry, Rheinisch-Westfälische Technische Hochschule Aachen University, 52074 Aachen, Germany; cDepartment ofChemistry, Roy & Diana Vagelos Laboratories, University of Pennsylvania, Philadelphia, PA 19104-6323; dInstitute of Computational Molecular Science,Temple University, Philadelphia, PA 19122; eNovioSense B.V., 6534 AT Nijmegen, The Netherlands; fDepartment of Cell and Developmental Biology,Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA 19104-6058; gDepartment of Bioengineering, University of Pennsylvania,Philadelphia, PA 19104-6321; hDepartment of Biology, University of Pennsylvania, PA 19104-6313; and iDepartment of Physics and Astronomy, University ofPennsylvania, Philadelphia, PA 19104-6059

Contributed by Michael L. Klein, January 22, 2019 (sent for review December 27, 2018; reviewed by Stephen Z. D. Cheng and Timothy J. Deming)

Self-assembling dendrimers have facilitated the discovery ofperiodic and quasiperiodic arrays of supramolecular architecturesand the diverse functions derived from them. Examples are liquidquasicrystals and their approximants plus helical columns andspheres, including some that disregard chirality. The same periodicand quasiperiodic arrays were subsequently found in blockcopolymers, surfactants, lipids, glycolipids, and other complexmolecules. Here we report the discovery of lamellar and hexago-nal periodic arrays on the surface of vesicles generated fromsequence-defined bicomponent monodisperse oligomers contain-ing lipid and glycolipid mimics. These vesicles, known as glyco-dendrimersomes, act as cell-membrane mimics with hierarchicalmorphologies resembling bicomponent rafts. These nanosegre-gated morphologies diminish sugar–sugar interactions enablingstronger binding to sugar-binding proteins than densely packedarrangements of sugars. Importantly, this provides a mechanismto encode the reactivity of sugars via their interaction with sugar-binding proteins. The observed sugar phase-separated hierarchicalarrays with lamellar and hexagonal morphologies that encode bi-ological recognition are among the most complex architectures yetdiscovered in soft matter. The enhanced reactivity of the sugardisplays likely has applications in material science and nanomedi-cine, with potential to evolve into related technologies.

Janus glycodendrimers | lipid rafts | nanosegregation | atomic forcemicroscopy | galectin

Soft matter self-organizes into a diversity of periodic andquasiperiodic arrays, including Frank–Kasper phases such as

cubic Pm�3n, known also as the A15 phase (1–4), tetragonal P42/mnm known the σ-phase (5), dodecagonal liquid quasicrystals(6–8), plus helical columns (9) and spheres (10), including somethat disregard chirality. The aforementioned morphologies areresponsible for a variety of functions in soft matter and biology(3, 11–15). These complex architectures, which were discoveredand elucidated via structural and retrostructural analysis of li-braries of self-assembling dendrons and dendrimers, employeddiffraction methods analogous to those used to develop the fieldof structural biology (16). Later these same phases were found bythe same methods (17) in block copolymers (18–22), surfactants(23–26), lipids (27–33), glycolipids (34), and in other systems(35–38). Recently, amphiphilic Janus dendrimers (JDs) (39, 40),and their sugar-presenting analogs, Janus glycodendrimers(JGDs) (41), which provide synthetic alternatives to naturallipids and glycolipids that are readily functionalized, have beenreported to self-assemble in either water or buffer (42) into vesicles,named dendrimersomes (DSs) and glycodendrimersomes (GDSs)(43), respectively. They can also be made into cell-like hybrids witheither bacterial (44) or human cells (45). As with cell membranes,

the bilayers of these supramolecular constructs act as a barrierbetween the inside and outside of DSs, GDSs, and cell-like hybrids(39, 40, 43, 44). Selective permeation can be provided, with com-parable efficiency as with liposomes (13) and polymersomes (46),when either transmembrane proteins or their mimics are incorpo-rated into the bilayer (44). Thus, DSs and GDSs can be used asbiological-membrane mimics to elucidate concepts in cell biologyand glycobiology, such as compartmentalization, encapsulation andrelease, selective transport (39, 40, 43, 44), as well as fusion andfission (47–49). Intriguingly, while searching for the minimumconcentration of sugar from JGDs that is able to aggregateproteins, we discovered that sequence-defined monodisperse JGDscontaining D-lactose (Lac) or D-mannose exhibit higher activity to-ward sugar-binding proteins, known as lectins, on decreasing surfacedensity of sugar (50–52). This inverse relationship between sugardensity and aggregation of lectins is opposite to what is expected

Significance

The seminal fluid mosaic model of the cell membranes suggestsa lipid bilayer sea, in which cholesterol, proteins, glycoconju-gates, and other components are swimming. Complementingthis view, a microsegregated rafts model predicts clusters ofcomponents that function as relay stations for intracellularsignaling and trafficking. However, elucidating the arrange-ment of glycoconjugates responsible for communication andrecognition between cells, and cells with proteins remainsa challenge. Herein, designed dendritic macromolecules areshown to self-assemble into vesicles that function as biological-membrane mimics with controlled density of sugar moieties ontheir periphery. Surprisingly, lowering sugar density elicitshigher bioactivity to sugar-binding proteins. This finding in-forms a design principle for active complex soft matter withpotential for applications in cellular biology and nanomedicine.

Author contributions: C.R.-E., Q.X., M.L.K., and V.P. designed research; C.R.-E., Q.X.,N.Y.K., S.E.S., K.R., B.E.P., S.L., D.S., A.M.R.P., I.B., H.H., M.K., and I.M. performed research;M.M., C.J.W., M.C.G., M.G., and T.B. contributed new reagents/analytic tools; C.R.-E., Q.X.,N.Y.K., B.E.P., M.L.K., and V.P. analyzed data; and C.R.-E., Q.X., N.Y.K., B.E.P., M.L.K., andV.P. wrote the paper.

Reviewers: S.Z.D.C., The University of Akron; and T.J.D., University of California,Los Angeles.

The authors declare no conflict of interest.

This open access article is distributed under Creative Commons Attribution-NonCommercial-NoDerivatives License 4.0 (CC BY-NC-ND).1To whom correspondence may be addressed. Email: mlklein@temple.edu or percec@sas.upenn.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1821924116/-/DCSupplemental.

Published online February 28, 2019.

5376–5382 | PNAS | March 19, 2019 | vol. 116 | no. 12 www.pnas.org/cgi/doi/10.1073/pnas.1821924116

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

12, 2

020

from classic kinetics (53, 54). Although the molecular structure ofthe JGDs exhibiting increased reactivity with decreasing sugardensity is known (50–52), the three-dimensional structure providingthis function is not understood. Indeed, elucidating the arrangementof glycoconjugates responsible for communication and recognitionbetween cells, and cells with proteins (49, 55), remains a majorchallenge. Progress toward this goal, which will also enable thedesign of glycoconjugate-based biomaterials, is reported here.Seminal models of the structure of the cell membrane, in-

cluding the lipid bilayer sea of the fluid mosaic model (56), andthe microsegregated rafts model (57, 58), recognize that thereare certain structures responsible for sugar (glycan) activity buthave not fully elucidated the details of these structures. This ismostly since X-ray and other diffraction methods employed inthe structural and retrostructural analysis of biological systemand of dendrimer assemblies cannot be easily applied to theanalysis of fluid natural and synthetic cell membranes (59–61).Here, we employ an analysis of atomic force microscopy im-

ages and their associated fast Fourier transforms (FFTs), alongwith complementary molecular modeling, to address the questionof what architectures program enhanced glycan function of JGDsvesicles. This diffraction-like methodology reveals that a complex (62,63) hierarchical self-organization of sugar moieties arranged into la-mellar and hexagonal nanophase-separated periodic arrays are re-sponsible for the enhanced activity compared with GDS where sugarswere densely packed and do not exhibit nanomorphologies. Suchdense packing presumably results in stronger sugar–sugar interac-tions ultimately reducing their availability for binding to galectins.Furthermore, these bicomponent patterned hierarchical nano-architectures may provide a general mechanism to program andencode biological recognition and function of glycan, not only onbicomponent GDSs, but also on the surface of different vesiclescontaining glycoconjugates such as liposomes, polymersomes, syn-thetic cells, hybrid cells, and other globular assemblies.

Results and DiscussionAccelerated Modular Synthesis of JGDs and Sequence-Defined OligomericJGDs. GDSs were self-assembled from sequence-defined JGDs (Figs.1 and 2A). All JGDs in this work contain Lac, a sugar commonlyfound in human cells, and triethylene glycol (3EO) fragments (Figs. 1and 2B) as their hydrophilic components. Lac substituents are dilutedby increasing the number of 3EO substituents to create a library withsystematic structural variation. The synthesis of JGD-4Lac is shown inSI Appendix, Figs. S1–S8, while that of JGD(5/1Lac

2) and JGD(5/1Lac3)

in SI Appendix, Figs. S9–S17.Other JGDs (Fig. 1) were prepared as previously reported by

accelerated iterative modular synthesis (39, 51), similar to JGD-4Lac. These JGDs share the same hydrophobic part containing3,5-bis(dodecyloxy)benzoic esters. A library including high-sugar-density single–single JGD-1Lac, twin–twin JGD-2Lac, and tetra–tetraJGD-4Lac, and sequence-defined JGDs with relatively low-sugardensity were involved in this study. The sugar densities decreasedfrom 100% Lac (JGD-1Lac, JGD-2Lac, and JGD-4Lac) to 3/1 (3EO/Lac) for JGD(3/1Lac), 5/1 (for 3EO/Lac) for JGD(5/1Lac

2) and JGD(5/1Lac

3), 6/1 (3EO/Lac) for JGD(6/1Lac), and 8/1 (3EO/Lac) for JGD(8/1Lac

2S), JGD(8/1Lac3S), JGD(8/1Lac

2L), and JGD(8/1Lac3L).

The position of Lac on the hydrophilic 3EO dendrons was tunedin the 8/1 (3EO/Lac) sequence-defined JGDs. Lac was appended inthe 2 position (counted from left to right) of the gallic amide forJGD(8/1Lac

2S) and JGD(8/1Lac2L), and in the 3 position for JGD(8/

1Lac3S) and JGD(8/1Lac

3L). The linker lengths of the sequence-defined JGDs were changed from a single 3EO unit (short linker,denoted as S) in JGD(8/1Lac

2S) and JGD(8/1Lac3S) to two 3EO units

(long linker, denoted as L) in JGD(8/1Lac2L) and JGD(8/1Lac

3L).

Hierarchical Periodic Glycan Arrays Self-Organized by Self-Assemblyof High-Density and Sequence-Defined JGDs. Lectins are proteinswhich recognize glycoconjugates on the cell membranes and thus

mediate communication between cells with cell and with pro-teins. GDSs self-assembled from libraries of JGDs unveiled howthe structures of lectins affect their bioactivity toward recogni-tion by investigating the cross-linking of cell-like GDSs and thefunctional pairing between the cognate sugars and lectins (64–68). Lac-presenting JGDs were selected for the present studysince they self-assemble into unilamellar vesicles comparativelysimpler than those of onion-like multilamellar vesicles self-assembled from Man-presenting JGDs and since Lac is largerthan Man and which is expected to facilitate analysis by atomicforce microscopy (AFM) experiments.Aggregation of GDSs with wild-type human galectin-1 (Gal-1)

(Fig. 2C) reproduces the unexpected but general trend of an in-crease in activity of GDSs with decreasing density of sugars reportedpreviously in both unilamellar (51) and onion-like (52) GDSs.There is a modest, yet statistically significant, increase in the activityfrom JGD-1Lac to JGD-2Lac, and to JGD-4Lac (Fig. 2C and SI Ap-pendix, Fig. S18), as expected by multivalency (53, 54). There is amarked increase in aggregation activity when the sugar density isdecreased, in sequence-defined JGD(5/1Lac

2), JGD(6/1Lac3), JGD(6/

1Lac), JGD(8/1Lac2S), JGD(8/1Lac

3S), JGD(8/1Lac2L), and JGD(8/

1Lac3L) (Fig. 2E and SI Appendix, Fig. S19). Similarly, other complex

galectins including heterodimeric human galectin-8S (Gal-8S) (50,51) and engineered galectin-1 tetramer (Gal-1)4–GG (67) presentedtrends in line with Gal-1 (Fig. 2C).

Structural Analysis of Lamellar Periodic Glycan Arrays of Giant GDSsSelf-Organized from High-Density and Sequence-Defined JGDs. Astructural analysis methodology using AFM, FFT, and molecularmodeling was used to investigate the structures of GDSs (Fig. 3).Giant GDSs prepared by thin-film hydration were utilized, as thedimensions of giant GDSs (micrometer scale) rather than thoseof small GDSs (100–200 nm) obtained by injection (39) are requiredto access analysis of hierarchical nanodomains by AFM. FFT pro-vides a diffraction-like pattern that is used for structural and ret-rostructural analysis. Giant GDSs prepared from JGD-1Lac andJGD-2Lac display a very smooth and compact surface (Fig. 4A) whilethose prepared from JGD-4Lac (Fig. 4B), JGD(3/1Lac), JGD(5/1Lac

2),JGD(6/1Lac

3), JGD(6/1Lac), JGD(8/1Lac2S), and JGD(8/1Lac

3S)(Fig. 4C) unexpectedly exhibit hierarchical lamellar morphologies.Additional sequence-defined JGDs in which Lac is attached via atwo-3EO linker JGD(8/1Lac

2L), and JGD(8/1Lac3L) (Fig. 2C) generate

GDSs which exhibit hierarchical periodic arrays of glycans with hex-agonal patterns (Fig. 4D). Although the lamellar and hexagonal pe-riodicities of these glycan arrays resemble those of supramoleculardendrimers (11, 16) and block copolymers (18–22), the three-dimensional structures of these GDSs represent an arrangement ofbuilding blocks on the surface of a vesicle. Detailed AFM images in-cluding phase and height profiles, and additional modeling of GDSsof self-assembled JGDs, are available in SI Appendix, Figs. S20–S25.What is the mechanism via which hierarchical periodic glycan

arrays with programmed activity are generated and how do theyinfluence reactivity to galectin? Homogeneous and denselypacked surfaces were observed only for single–single JGD-1Lacand twin–twin JGD-2Lac (Figs. 1 and 2A) that can self-organizeinto GDSs with high-density bilayers (Fig. 4A). In contrast, tetra–tetra JGD-4Lac and all sequence-defined JGDs present morecomplex, hierarchically self-organized nanodomain patterns (Fig.4 B–D) with lamellar and hexagonal morphologies. The nano-segregation of sequence-defined JGDs may be anticipated sincethey are generated from dissimilar hydrophilic chemical frag-ments. However, the formation of nanosegregated domains byJGD-4Lac is unexpected and it is presumably driven by a differentmechanism than the one acting in the case of sugar dilution.Molecular modeling (Fig. 5 A and B) suggests that Lac moieties seekto maximize their interactions with each other. This can be achievedwithout nanosegregation in densely packed JGD-1Lac and JGD-2Lac(Fig. 4B). In contrast, the sugar moieties of JGD-4Lac must cluster

Rodriguez-Emmenegger et al. PNAS | March 19, 2019 | vol. 116 | no. 12 | 5377

CHEM

ISTR

YCE

LLBIOLO

GY

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

12, 2

020

together to maximize interactions, resulting in alternating high- andlow-density regions of Lac that exhibit a lamellar periodic array(Figs. 4B and 5 A and B). The change from JGD-1Lac and JGD-2Lacto JGD-4Lac resembles the classic polymer effect that provides atransition from rigid to flexible backbone conformation (3).Sequence-defined JGD(5/1Lac

2), JGD(5/1Lac3), JGD(6/1Lac),

JGD(8/1Lac2S), and JGD(8/1Lac

3S) with programmed enhanced ac-tivity generate similar periodic lamellar structures to JGD-4Lac, exceptthat both the sugar density and the steric interactions are lower thanin JGD-4Lac (Fig. 4 B andC and SI Appendix, Fig. S23). Consequently,the clusters of sugars on the surface of the GDSs self-assembled fromsequence-defined JGDs are more accessible to lectin and theiraggregation is therefore more efficient than for JGD-4Lac eventhough having the same lamellar morphology (Fig. 2C).

Structural Analysis of Hexagonal Periodic Glycan Arrays of Giant GDSsSelf-Organized from Sequence-Defined JGDs. In contrast, JGD(8/1Lac

2L) and JGD(8/1Lac3L) exhibit periodic hexagonal arrays (Fig.

4D), as demonstrated by indexing the FFT of their AFM phaseimages (shown for JGD(8/1Lac

3L) in Fig. 5C). This hexagonalperiodic array provides an alternative morphology to enhancesugar activity. AFM height analysis (Fig. 5D, Right) shows thatthe isolated domains are thinner (darker in Fig. 5D) comparedwith the continuous domain (lighter in Fig. 5D). This suggeststhat the Lac moieties form the continuous domain.A model of hexagonal nanosegregation consistent with anal-

ysis of the FFT of AFM images (Fig. 5 E and F) comprises19 molecules of JGD(8/1Lac

3L): Lac moieties [red, space-fillingCorey–Pauling–Koltun (CPK) view] interact at the edge of thedomain, while the hydrophilic (blue) and hydrophobic (green)components of JGD(8/1Lac

3L) occupy the remainder of the do-main in three concentric layers: an outer layer (orange circles, I);an inner layer (purple circles, II); and a molecule at the center ofthe column (yellow circle, III). The hydrophilic and hydrophobiccomponents of one molecule from each concentric layer aredepicted in CPK view, while those of the other molecules are

Fig. 1. Molecular structures of JGDs. Structures and short nomenclatures are shown for (Top) high-sugar density JGD-1Lac, JGD-2Lac, and JGD-4Lac, andsequence-defined JGD(3/1Lac), JGD(5/1Lac

2), JGD(5/1Lac3), JGD(6/1Lac); (Bottom) JGD(8/1Lac

2S), JGD(8/1Lac3S), JGD(8/1Lac

2L), and JGD(8/1Lac3L).

5378 | www.pnas.org/cgi/doi/10.1073/pnas.1821924116 Rodriguez-Emmenegger et al.

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

12, 2

020

shown as faint line models for clarity. The linker of JGD(8/1Lac3L)

molecules contains two 3EO units (Figs. 1 and 2). In moleculesforming the outer layer (I), this linker is disordered beneath Lac.For molecules in the inner layer (II), the linker is more extendedto allow the Lac moiety to interact with sugars from molecules inthe outer layer (I). Finally, the linker of the central molecule ofJGD(8/1Lac

3L) (III) is completely extended to allow interactionbetween sugars. A molecule at the center of a larger domainwould be unable to position its Lac moiety to allow for in-teraction with Lac moieties from the outer (I) and inner (II)layers, and hence the linker may restrict the size of the periodichexagonal array. Conversely, the length of the two-3EO linkermay be too long to allow sufficient confinement in periodic la-mellar domains, consistent with the observation of lamellarphases only for sequence-defined JGDs with shorter single-3EOlinkers, JGD(3/1Lac), JGD(6/1Lac), JGD(8/1Lac

2S), and JGD(8/1Lac

3S) (Fig. 4D and SI Appendix, Fig. S24A). On the other hand,two sequence-defined JGDs JGD(5/1Lac

2) and JGD(5/1Lac3) with

two-3EO display only lamellar domains due to higher sugardensity compared with JGD(8/1Lac

3L). These data, along with thegeneration of a hexagonal periodic array by a second moleculewith a two-3EO linker, JGD(8/1Lac

2L) (SI Appendix, Fig. S24B),suggest that both the linker length and sugar density determinethe observed morphology. A columnar section of the bilayercomprising the repeating unit of the periodic hexagonal array(Fig. 5F) illustrates how hierarchical assembly of JGD(8/1Lac

3L)(Fig. 2A) containing a nonaggregating single Lac becomes a veryefficient ligand for aggregation with lectins. Both the hexagonaland the nondense lamellar arrays from Figs. 4 and 5 indicatepotential mechanisms, which by steric constrain–release com-bined with self-organization (Fig. 4B) and/or only self-organization(Figs. 4 C and D and 5F), generate more reactive glycan ligandsthan those obtained from highly packed assemblies (Fig. 4A).

The clustering of sugars by self-organization into lamellar andhexagonal periodic arrays reported here suggests a mechanistic ex-planation for the enhanced reactivity observed with sequence-defined JGDs that was not predicted by the rafts model (57, 58).

Encoding Biological Recognition by Hierarchical Nanomorphology.The goal of this paper was to report the discovery of nano-periodic arrays on vesicles generated from sequence-definedbicomponent monodisperse oligomers containing lipid and gly-colipid mimics and illustrate how they encode biological recognitionand reactivity. Remarkably these nanosegregated morphologieswith diminished sugar–sugar aggregation allowed the sugar tomore readily bind to galectins than the densely packed arrange-ments of sugars, regardless of sugar dilution. The activity of thesenanosegregated glycans depends on both morphology and sugardensity (Fig. 2C). For lamellar morphology the binding to galectinsincreased with the decreasing density of sugar until a 1:6 dilutionafter which no further significant increase was observed (Fig. 2Cand SI Appendix, Table S1). No statistically significant differenceswere observed for the binding between JGD(8/1Lac

2S) and JGD(8/1Lac

3S) displaying lamellar or between JGD(8/1Lac2L) and JGD(8/

1Lac3L) displaying hexagonal morphology. However, a comparison

between these two morphologies at the same dilution, with statis-tically significant results, evidenced higher activity for the lamellar,JGD(8/1Lac

2S) and JGD(8/1Lac3S), rather than for the hexagonal,

JGD(8/1Lac2L) and JGD(8/1Lac

3L), organization (Fig. 2C and SIAppendix, Table S1) with Gal-1 and Gal-8S.

ConclusionsThe structures, morphologies, and mechanisms via which the bi-component sequence-defined GDSs with decreasing sugar con-centration provide enhanced glycan activity have been elucidatedwith a diffraction-like analysis methodology. The cell-membrane

Fig. 2. Molecular architecture JGDs containing defined glycan sequence and density. (A) Sequence-defined JGDs with different Lac density, sequence, andlinker length. (B) Schematic representation of JGD building blocks. (C) Summary of aggregation assay data using GDSs from self-assembly of sequence-definedJGDs (Lac = 0.1 mM, 900 μL) with Gal-1 (1 mg·mL−1, 100 μL), Gal-8S (1 mg·mL−1, 100 μL), and (Gal-1)4–GG (1 mg·mL−1, 100 μL). Color codes for galectins: Gal-1,red; Gal-8S, blue; (Gal-1)4–GG, green. N and C represent the N and C termini of proteins. For selected examples symbols used for significant difference (Pvalues by Student’s t test) are: “n.s.” for P > 0.05 (for statistically nonsignificant) and “*” for P < 0.05 (for statistically significant).

Rodriguez-Emmenegger et al. PNAS | March 19, 2019 | vol. 116 | no. 12 | 5379

CHEM

ISTR

YCE

LLBIOLO

GY

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

12, 2

020

mimic, GDSs assembled from JGD-1Lac and JGD-2Lac (Fig. 2A),display a very compact arrangement of Lac units on their surface(Fig. 4A). GDSs programmed to exhibit higher activity by self-assembly of sequence-defined JGDs (Figs. 2A and 4 B–D) self-organize hierarchical periodic arrays of glycans that reduce thedense packing of Lac. Favorable interactions with sugar-bindingproteins are generated by these morphologies, resulting in an in-creased rate constant for aggregation, which overcomes the de-crease in the concentration of the sugar on the GDS surface. Thisfinding seemingly contradicts classic kinetic principles (53, 54), butis consistent with previous studies of self-assembled monolayers ofsugars on gold surfaces (69), which demonstrated that lower sugardensity provides higher affinity when applied as biosensors forlectins and viruses. However, in those studies (69), the arrange-ment of sugars within the monolayer was not elucidated. The hi-erarchical segregation of JGDs that have identical hydrophobicsegments differ only in their hydrophilic parts for the self-assemblyof vesicles. Neither the fluid mosaic model of the structure of cellmembranes (56) nor the rafts membrane-organizing principles(57, 58) predict these glycan nanodomain morphologies on thestructure of biological cell membranes or even on the structure of

biological-membrane mimics. The observed hierarchical nano-domains represent an example of bicomponent rafts-like mor-phology that extends from a small rafts domain to the entirebiological-membrane mimic and therefore could be elucidated atthe molecular level. The hierarchical morphologies reported hereare general for all unilamellar GDS generated by hydration andinjection (41) regardless of the size of the carbohydrate from JGDand also for multilamellar “onion-like” (52) GDS. Quantitativecorrelations between hierarchical morphology and sugar reactivityrequires the development of methodologies accessible below thelimit of the diffraction of light. At the same time, these patternednanoarchitectures are among the most complex structures yetdiscovered either on a planar or nonplanar surface (70–72). Theyprovide a powerful example in which structure determines func-tion (3, 10–14, 73), in this particular example how different su-pramolecular assemblies encode biological recognition. Thissupramolecular patterning offers a fundamental mechanism forprogramming enhanced activity via the surface morphology ofglobular glycoconjugate assemblies, such as liposomes, polymer-somes, and cell-like hybrids (44, 45). Indeed, this approach in-forms a general design and structural analysis principle for activecomplex soft matter (62, 63), which will enable technological ap-plications ranging from materials science to nanomedicine (74). Itremains to be seen if bottom-up approaches derived from FFT ofAFM (75) images elaborated here can overcome the limits ofother methods employed in the analysis of rafts (59–61, 76, 77).

MethodsPreparation of Nanoscale GDSs by Injection. A stock solution was prepared bydissolving the required amount of amphiphilic JGDs in THF. GDSs were thengenerated by injection of 50 μL of the stock solution into 1.0 mL PBS, fol-lowed by 5-s vortexing.

Preparation of Giant GDSs by Hydration. A solution of JGDs in THF (100 μL) wasdeposited on the top surface of a roughened Teflon sheet (1 cm2), placed ina flat-bottom vial, followed by evaporation of the solvent for 2 h. The Teflonsheet was dried in vacuo for an additional 12 h. Milli-Q water (1.0 mL) wasadded to submerge the film on the Teflon sheet, and the vial was placed in a60 °C oven for 12 h for hydration. The sample was then mixed using a vortexmixer for 30 s with a final concentration of 1 mg·mL−1.

Fig. 3. Illustration of combined AFM, FFT, and modeling methodology forstructural analysis.

Fig. 4. Surface topology of GDS formed by self-organization of JGDs depends on glycan sequence and density. (A) JGD-2Lac; (B) JGD-4Lac; (C) JGD(6/1Lac); (D)JGD(8/1Lac

3L). (B–D, Inset) FFT of AFM phase images. Color code: red, sugars; blue, hydrophilic 3EO; green, hydrophobic alkyl.

5380 | www.pnas.org/cgi/doi/10.1073/pnas.1821924116 Rodriguez-Emmenegger et al.

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

12, 2

020

AFM. The GDSs in water (∼0.3 mg·mL−1) were deposited and slowly dried onthe freshly peeled mica. All images were acquired with a Multimode AFMNanoScope V (Digital Instruments) as topological scans in tapping mode inair, using silicon probes OTESPA-R3 (Bruker) with a nominal spring constantof 26 N·m−1 and a tip radius of 7 nm. The phase images were obtained bymonitoring the phase lag of the cantilever vibration compared with the z-piezo-drive voltage while the probe scans the surface with a preset constantamplitude of vibration. The phase data contain additional informationabout the tip–sample interactions resulting from adhesion, surface stiffness,and viscoelastic effects. The AFM scans including FFT were analyzed usingGwyddion software.

Cryogenic Transmission Electron Microscopy. Cryogenic transmission electronmicroscopy (cryo-TEM) micrographs were taken on a Carl Zeiss Libra120 microscope. Cryo-TEM samples were prepared by plunge freezing ofaqueous dispersion on plasma-treated lacey grids. The vitrified specimenswere transferred to a Gatan-910 cryoholder. The images were recorded at atemperature of −170 °C with an acceleration voltage of 120 kV. The GDSvesicle structures of JGD-1Lac, JGD-2Lac, JGD(3/1Lac), JGD(6/1Lac), JGD(8/1Lac

2S),JGD(8/1Lac

3S), JGD(8/1Lac2L), and JGD(8/1Lac

3L) were synthesized and charac-terized as reported previously (51, 65). Additional cryo-TEM images are pre-sented in SI Appendix, Fig. S26.

Modeling of Bilayers and Vesicles. Models of bilayers in Figs. 4 and 5 and SIAppendix, Fig. S25 are idealized depictions of the arrangement of JDs andGSDs within a vesicle, consistent with height and phase profiles obtained byAFM. The energy of molecular structures was minimized using Chem3D(MM2 module), and then the torsional angles of bonds at the core of thedendrimer were modified in Accelrys Discovery Studio to provide segre-gation between the hydrophilic and hydrophobic portions of the mole-cule. Assemblies were generated by manual placement of molecules inAccelrys DS ViewerPro 5.0. The height profile of the assembly model was

calculated and compared against experimental data obtained from AFM.The model was iteratively refined until good agreement between calcu-lated height profile and experimental height or phase profile wasobtained.

Models of vesicles in Fig. 4 are schematic depictions of the general to-pology of the surface of the GDS vesicle and are not to scale. These modelswere prepared using Autodesk 3DS Max 2017.

Aggregation Assays. Aggregation assays of GDSs with lectins were monitoredin semimicrodisposable cuvettes (path length, l = 0.23 cm) at 23 °C atwavelength λ = 450 nm by using a Shimadzu UV-vis spectrophotometer UV-1601 with Shimadzu/UV Probe software in kinetic mode. PBS solution ofgalectin (100 μL) was injected into PBS solution of GDSs (900 μL). The cuvettewas shaken by hand for 1–2 s before data collection was started. The samesolution of GDSs solution was used as a reference. PBS solutions of galectinwere prepared before the agglutination assays and were maintained at 0 °C(ice bath) before data collection.

ACKNOWLEDGMENTS. The authors thank Professor Hans-Joachim Gabius ofLudwig Maximilian University, Munich for providing lectin samples and fordiscussions; Dr. Mihai Peterca for discussions; and the two reviewers forconstructive suggestions and recommendations. This work was supported byNational Science Foundation Grants DMR-1066116 and DMR-1807127 (toV.P.); the P. Roy Vagelos Chair at the University of Pennsylvania (V.P.); theAlexander von Humboldt Foundation (N.Y.K. and V.P.); National ScienceFoundation Grants DMR-1120901 (to M.L.K., M.G., and V.P.) and DMR-1720530 (to M.G. and V.P.); National Institutes of Health Grant R01-GM080279 (to M.G.); Howard Hughes Medical Institute for an InternationalStudent Research Fellowship (to B.E.P.); the European Union’s Horizon2020 research and innovation programme under the Marie Skłodowska-Curie Grant Agreement 642687 (to I.B.); and financial support from NankaiUniversity (H.H.).

Fig. 5. Models of nanosegregated bilayer structures. (A and B) Models of JGD-4Lac: (A) single sugar cluster comprising three molecules on each side ofthe bilayer; (B) sugar clusters in bilayer. (C) FFT of the AFM phase image of JGD(8/1Lac

3L) (Fig. 4D) shows multiple features consistent with formation of a hex-agonal array. Blue, yellow, and green circles indicate (10), (11), and (20) features, respectively. The theoretical ratio a:b:c for a hexagonal array is 1:√3(=1.73):2; the observed ratio a:b:c is 1:1.75:2.04. (D, Left) Model of hexagonal nanosegregation and (D, Right) AFM height image of JGD(8/1Lac

3L).(E ) Top view of bilayer model with all Lac moieties and three highlighted JGD(8/1Lac

3L) molecules in CPK view. (F ) Side view of a column of the bilayer ofJGD(8/1Lac

3L).

Rodriguez-Emmenegger et al. PNAS | March 19, 2019 | vol. 116 | no. 12 | 5381

CHEM

ISTR

YCE

LLBIOLO

GY

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

12, 2

020

1. Balagurusamy VSK, Ungar G, Percec V, Johansson G (1997) Rational design of the firstspherical supramolecular dendrimers self-organized in a novel thermotropic cubicliquid-crystalline phase and the determination of their shape by X-ray analysis. J AmChem Soc 119:1539–1555.

2. Hudson SD, et al. (1997) Direct visualization of individual cylindrical and sphericalsupramolecular dendrimers. Science 278:449–452.

3. Percec V, et al. (1998) Controlling polymer shape through the self-assembly of den-dritic side-groups. Nature 391:161–164.

4. Dukeson DR, et al. (2003) Application of isomorphous replacement in the structuredetermination of a cubic liquid crystal phase and location of counterions. J Am ChemSoc 125:15974–15980.

5. Ungar G, Liu Y, Zeng X, Percec V, Cho W-D (2003) Giant supramolecular liquid crystallattice. Science 299:1208–1211.

6. Zeng X, et al. (2004) Supramolecular dendritic liquid quasicrystals. Nature 428:157–160.7. Ungar G, Percec V, Zeng X, Leowanawat P (2011) Liquid quasicrystals. Isr J Chem 51:1206–1215.8. Holerca MN, et al. (2018) Dendronized poly(2-oxazoline) displays within only five

monomer repeat units liquid quasicrystal, A15 and σ Frank–Kasper phases. J Am ChemSoc 140:16941–16947.

9. Roche C, et al. (2014) Homochiral columns constructed by chiral self-sorting during su-pramolecular helical organization of hat-shapedmolecules. J Am Chem Soc 136:7169–7185.

10. Percec V, et al. (2009) Self-assembly of dendronized triphenylenes into helical pyra-midal columns and chiral spheres. J Am Chem Soc 131:7662–7677.

11. Sun H-J, Zhang S, Percec V (2015) From structure to function via complex supramo-lecular dendrimer systems. Chem Soc Rev 44:3900–3923.

12. Percec V, et al. (2002) Self-organization of supramolecular helical dendrimers intocomplex electronic materials. Nature 419:384–387.

13. Percec V, et al. (2004) Self-assembly of amphiphilic dendritic dipeptides into helicalpores. Nature 430:764–768.

14. Percec V, Rudick JG, Peterca M, Heiney PA (2008) Nanomechanical function from self-organizable dendronized helical polyphenylacetylenes. J Am Chem Soc 130:7503–7508.

15. Feringa BL, Browne WR (2008) Nanomechanics: Macromolecules flex their muscles.Nat Nanotechnol 3:383–384.

16. Rosen BM, et al. (2009) Dendron-mediated self-assembly, disassembly, and self-organization of complex systems. Chem Rev 109:6275–6540.

17. Peterca M, Percec V (2010) Materials science. Recasting metal alloy phases with blockcopolymers. Science 330:333–334.

18. Lee S, Bluemle MJ, Bates FS (2010) Discovery of a Frank-Kasper σ phase in sphere-forming block copolymer melts. Science 330:349–353.

19. Bates FS, et al. (2012) Multiblock polymers: Panacea or Pandora’s box? Science 336:434–440.20. Gillard TM, Lee S, Bates FS (2016) Dodecagonal quasicrystalline order in a diblock

copolymer melt. Proc Natl Acad Sci USA 113:5167–5172.21. Chanpuriya S, et al. (2016) Cornucopia of nanoscale ordered phases in sphere-forming

tetrablock terpolymers. ACS Nano 10:4961–4972.22. Kim K, et al. (2017) Thermal processing of diblock copolymer melts mimics metallurgy.

Science 356:520–523.23. Perroni DV, Mahanthappa MK (2013) Inverse Pm3n cubic micellar lyotropic phases

from zwitterionic triazolium gemini surfactants. Soft Matter 9:7919–7922.24. Yue K, et al. (2016) Geometry induced sequence of nanoscale Frank-Kasper and

quasicrystal mesophases in giant surfactants. Proc Natl Acad Sci USA 113:14195–14200.25. Kim SA, Jeong K-J, Yethiraj A, Mahanthappa MK (2017) Low-symmetry sphere

packings of simple surfactant micelles induced by ionic sphericity. Proc Natl Acad SciUSA 114:4072–4077.

26. Baez-Cotto CM, Mahanthappa MK (2018) Micellar mimicry of intermetallic C14 andC15 Laves phases by aqueous lyotropic self-assembly. ACS Nano 12:3226–3234.

27. Mariani P, Luzzati V, Delacroix H (1988) Cubic phases of lipid-containing systems.Structure analysis and biological implications. J Mol Biol 204:165–189.

28. Vargas R, Mariani P, Gulik A, Luzzati V (1992) Cubic phases of lipid-containing sys-tems. The structure of phase Q223 (space group Pm3n). An X-ray scattering study.J Mol Biol 225:137–145.

29. Sakya P, Seddon JM, Templer RH, Mirkin RJ, Tiddy GJT (1997) Micellar cubic phasesand their structural relationships: The nonionic surfactant system C12EO12/water.Langmuir 13:3706–3714.

30. Charvolin J, Sadoc JF (1988) Periodic systems of frustrated fluid films and “micellar”cubic structures in liquid crystals. J Phys 49:521–526.

31. Paccamiccio L, et al. (2006) Pressure effects on lipidic direct phases: The dodecyl tri-methyl ammonium chloride-water system. J Phys Chem B 110:12410–12418.

32. Bastos M, et al. (2011) Lactoferrin-derived antimicrobial peptide induces a micellarcubic phase in a model membrane system. Biophys J 101:L20–L22.

33. Silva T, et al. (2013) Structural diversity and mode of action on lipid membranes of threelactoferrin candidacidal peptides. Biochim Biophys Acta 1828:1329–1339.

34. Gulik A, Delacroix H, Kirschner G, Luzzati V (1995) Polymorphism of ganglioside-water systems: A new class of micellar cubic phases. Freeze-fracture electron mi-croscopy and X-ray scattering studies. J Phys II 5:445–464.

35. Huang M, et al. (2015) Self-assembly. Selective assemblies of giant tetrahedra viaprecisely controlled positional interactions. Science 348:424–428.

36. Zhang W, et al. (2016) Toward controlled hierarchical heterogeneities in giant mol-ecules with precisely arranged nano building blocks. ACS Cent Sci 2:48–54.

37. Zhang W, et al. (2017) Sequence-mandated, distinct assembly of giant molecules.Angew Chem Int Ed Engl 56:15014–15019.

38. Ye X, et al. (2017) Quasicrystalline nanocrystal superlattice with partial matchingrules. Nat Mater 16:214–219.

39. Percec V, et al. (2010) Self-assembly of Janus dendrimers into uniform den-drimersomes and other complex architectures. Science 328:1009–1014.

40. Peterca M, Percec V, Leowanawat P, Bertin A (2011) Predicting the size and propertiesof dendrimersomes from the lamellar structure of their amphiphilic Janus den-drimers. J Am Chem Soc 133:20507–20520.

41. Percec V, et al. (2013) Modular synthesis of amphiphilic Janus glycodendrimers andtheir self-assembly into glycodendrimersomes and other complex architectures withbioactivity to biomedically relevant lectins. J Am Chem Soc 135:9055–9077.

42. Ball P (2008) Water as an active constituent in cell biology. Chem Rev 108:74–108.43. Sherman SE, Xiao Q, Percec V (2017) Mimicking complex biological membranes and

their programmable glycan ligands with dendrimersomes and glycodendrimersomes.Chem Rev 117:6538–6631.

44. Xiao Q, et al. (2016) Bioactive cell-like hybrids coassembled from (glyco)dendrimersomeswith bacterial membranes. Proc Natl Acad Sci USA 113:E1134–E1141.

45. Yadavalli SS, et al. (2019) Bioactive cell-like hybrids from dendrimersomes with ahuman cell membrane and its components. Proc Natl Acad Sci USA 116:744–752.

46. Kaucher MS, et al. (2007) Selective transport of water mediated by porous dendriticdipeptides. J Am Chem Soc 129:11698–11699.

47. Xiao Q, et al. (2016) Self-sorting and coassembly of fluorinated, hydrogenated, andhybrid Janus dendrimers into dendrimersomes. J Am Chem Soc 138:12655–12663.

48. Xiao Q, et al. (2017) Janus dendrimersomes coassembled from fluorinated, hydro-genated, and hybrid Janus dendrimers as models for cell fusion and fission. Proc NatlAcad Sci USA 114:E7045–E7053.

49. Bertozzi CR, Kiessling LL (2001) Chemical glycobiology. Science 291:2357–2364.50. Zhang S, et al. (2015) Dissecting molecular aspects of cell interactions using glyco-

dendrimersomes with programmable glycan presentation and engineered humanlectins. Angew Chem Int Ed Engl 54:4036–4040.

51. Zhang S, et al. (2015) Glycodendrimersomes from sequence-defined Janus glyco-dendrimers reveal high activity and sensor capacity for the agglutination by naturalvariants of human lectins. J Am Chem Soc 137:13334–13344.

52. Xiao Q, et al. (2016) Onion-like glycodendrimersomes from sequence-defined Janusglycodendrimers and influence of architecture on reactivity to a lectin. Proc Natl AcadSci USA 113:1162–1167.

53. Fasting C, et al. (2012) Multivalency as a chemical organization and action principle.Angew Chem Int Ed Engl 51:10472–10498.

54. Munoz EM, Correa J, Riguera R, Fernandez-Megia E (2013) Real-time evaluation ofbinding mechanisms in multivalent interactions: A surface plasmon resonance kineticapproach. J Am Chem Soc 135:5966–5969.

55. Seeberger PH, Werz DB (2007) Synthesis and medical applications of oligosaccharides.Nature 446:1046–1051.

56. Singer SJ, Nicolson GL (1972) The fluid mosaic model of the structure of cell mem-branes. Science 175:720–731.

57. Lingwood D, Simons K (2010) Lipid rafts as a membrane-organizing principle. Science327:46–50.

58. Sezgin E, Levental I, Mayor S, Eggeling C (2017) The mystery of membrane organization:Composition, regulation and roles of lipid rafts. Nat Rev Mol Cell Biol 18:361–374.

59. Heberle FA, et al. (2013) Bilayer thickness mismatch controls domain size in modelmembranes. J Am Chem Soc 135:6853–6859.

60. Nickels JD, et al. (2015) Mechanical properties of nanoscopic lipid domains. J AmChem Soc 137:15772–15780.

61. Nickels JD, et al. (2017) The in vivo structure of biological membranes and evidencefor lipid domains. PLoS Biol 15:e2002214.

62. Lehn J-M (2002) Toward self-organization and complex matter. Science 295:2400–2403.63. Lehn J-M (2013) Perspectives in chemistry–Steps towards complex matter. Angew

Chem Int Ed Engl 52:2836–2850.64. Xiao Q, et al. (2016) Why do membranes of some unhealthy cells adopt a cubic ar-

chitecture? ACS Cent Sci 2:943–953.65. Zhang S, et al. (2014) Mimicking biological membranes with programmable glycan

ligands self-assembled from amphiphilic Janus glycodendrimers. Angew Chem Int EdEngl 53:10899–10903.

66. Zhang S, et al. (2015) Unraveling functional significance of natural variations of ahuman galectin by glycodendrimersomes with programmable glycan surface. ProcNatl Acad Sci USA 112:5585–5590.

67. Kopitz J, et al. (2017) Reaction of a programmable glycan presentation of glyco-dendrimersomes and cells with engineered human lectins to show the sugarfunctionality of the cell surface. Angew Chem Int Ed Engl 56:14677–14681.

68. Xiao Q, et al. (2018) Exploring functional pairing between surface glycoconjugatesand human galectins using programmable glycodendrimersomes. Proc Natl Acad SciUSA 115:E2509–E2518.

69. Yeung SY, et al. (2017) Reversible self-assembled monolayers (rSAMs): Adaptablesurfaces for enhanced multivalent interactions and ultrasensitive virus detection. ACSCent Sci 3:1198–1207.

70. Schulz M, Binder WH (2015) Mixed hybrid lipid/polymer vesicles as a novel membraneplatform. Macromol Rapid Commun 36:2031–2041.

71. Vatankhah-Varnosfaderani M, et al. (2018) Chameleon-like elastomers with molecu-larly encoded strain-adaptive stiffening and coloration. Science 359:1509–1513.

72. Damasceno PF, Engel M, Glotzer SC (2012) Predictive self-assembly of polyhedra intocomplex structures. Science 337:453–457.

73. Kaplan CN, et al. (2017) Controlled growth and form of precipitating microsculptures.Science 355:1395–1399.

74. Farokhzad OC, Langer R (2009) Impact of nanotechnology on drug delivery. ACS Nano3:16–20.

75. Morandat S, Azouzi S, Beauvais E, Mastouri A, El Kirat K (2013) Atomic force mi-croscopy of model lipid membranes. Anal Bioanal Chem 405:1445–1461.

76. Baumgart T, Hess ST, Webb WW (2003) Imaging coexisting fluid domains in bio-membrane models coupling curvature and line tension. Nature 425:821–824.

77. Cebecauer M, et al. (2018) Membrane lipid nanodomains. Chem Rev 118:11259–11297.

5382 | www.pnas.org/cgi/doi/10.1073/pnas.1821924116 Rodriguez-Emmenegger et al.

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

12, 2

020